Ryosuke Okuta
プロジェクトマネージャー at wizOnChain株式会社
Chūō, Japan
Summary
🤩
Rockstar🎓
Top SchoolRyosuke Okuta is a technology leader and hands-on engineer with 11 years of experience, best known for helping build Chainer and CuPy at Preferred Networks where his GPU and numerical-optimization work contributed to projects that earned thousands of GitHub stars and wide research adoption. As former CTO and Director at PFN he led deployment of deep learning across industries, managed engineering organizations of roughly 100 people, and shaped hiring and evaluation systems that scaled the company. He now advises and engineers for multiple startups while serving as Project Manager at Sakana AI and CIO at Fixstars Investment, focusing recently on generative AI and LLM applications. His contributions range from low-level CUDA kernel optimization to productizing ML for industrial partners like FANUC, blending deep technical expertise with business and organizational strategy. An interesting, less obvious strength is his track record of improving performance and multi-GPU robustness in core libraries (CuPy) that underpin many downstream ML projects.
11 years of coding experience
修士, 情報科学研究科 システム情報科学専攻, 修士, 情報科学研究科 システム情報科学専攻 at 東北大学準学士, 電気情報工学科, 準学士, 電気情報工学科 at 一関工業高等専門学校
Github Skills (22)
gpgpu10
python10
image-processing10
chainer10
machine-learning10
numpy10
kernel10
gpu10
performance-optimization10
optimisation10
cupy10
accelerated-computing10
cuda10
computer-vision10
optimization10
Programming languages (7)
Github contributions (5)
cupy/cupy
Jun 2016 - Mar 2022
NumPy & SciPy for GPU
Role in this project:
Backend Developer
Backend DeveloperContributions:6 reviews, 1528 commits, 991 PRs in 5 years 10 months
Contributions summary:Ryosuke made several contributions, primarily focused on improving the CuPy library. The commits involve removing comments, fixing documentation, and integrating shape placeholders in the examples. Additionally, the commits include crucial fixes to the implementation of the Linear and Convolution2D classes and their interplay with multi-GPU environments, indicating expertise in core library development and CUDA-accelerated computing.
cudapythoncusolvergpunumpy
chainer/chainer
May 2015 - Apr 2019
A flexible framework of neural networks for deep learning
Role in this project:ML Engineer & Python Developer
ML Engineer & Python DeveloperContributions:1 release, 2156 commits, 1449 PRs in 4 years
Contributions summary:Ryosuke primarily contributed to the CuPy library, which is a NumPy-compatible array library for GPU acceleration. Their commits focused on improving the performance of the library, particularly within the context of deep learning frameworks like Chainer. The user's work involved optimizing and refactoring existing code, including the creation of new kernels and improving existing ones, related to core functions within the CuPy library, demonstrating a focus on numerical computation optimization.
cudapythonmxnetcaffe2flexible-framework
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.